How to make UGC ads with Seedance.

To make a UGC-style ad with Seedance, write the hook and script first. Generate a presenter, take a clean product photo, then animate both with image-to-video, one short shot at a time. Add the voice and check the lip sync, cut 9:16, 1:1 and 4:5 versions with captions, review every frame and label the ad as AI.

By Babar Ali, founder · Published 15 November 2026 · Sources checked 15 November 2026

On this page: Seedance · Workflow · Prompts · Shot list · Checks · Rules · Cost · FAQ

What is Seedance?

Seedance is a family of AI video models from ByteDance’s Seed team. It turns text, images, video and audio into short clips with sound. Seedance 2.0 was announced on 12 February 2026[1], and Seedance 2.5 on 31 July 2026, rolling out on Jimeng AI, Doubao Pro and other platforms[2]. BytePlus also lists cheaper 2.0 Fast and 2.0 Mini versions, and a 2.5 Flash from 15 November 2026[3][4].

Where you can use it

  • Dreamina, CapCut’s AI creation app, which began rolling 2.5 out to subscribers in late October 2026[5].
  • CapCut, which added Seedance 2.0 from March 2026[6] and sells 2.5 in its plans[7].
  • BytePlus ModelArk, the API for developers, billed per token[3].
  • Third-party platforms such as Higgsfield[8] and Replicate[9].

Availability has spread country by country, so check that it is live where you are[6][5].

The specs that matter for a UGC ad

BytePlus’s comparison of the two versions on its API, as of 15 November 2026[4].

Seedance 2.5 and 2.0 on the BytePlus API, as of 15 November 2026
SpecSeedance 2.5Seedance 2.0
Length per generation4 to 30 seconds4 to 15 seconds
Resolution480p, 720p, 1080p480p, 720p, 1080p, 4K
Aspect ratios21:9, 16:9, 4:3, 1:1, 3:4, 9:16 or adaptiveThe same
SoundGenerated with the picture, 11 languages listedGenerated with the picture
Image inputsA first frame, first and last frames, or up to 30 reference imagesThe same, with up to 9 reference images

The limits that shape a UGC ad

  • No real face as an upload: use a preset digital character, a face from a trusted model, or a person who has verified and authorized their files[10].
  • On-screen text can come out misspelled[11], so we add every word of text in the edit.
  • 4:5 is not a preset ratio: pass a 4:5 first frame with the ratio set to adaptive, or crop it from a 9:16 master, as we do[4].
  • ByteDance says complex motion and scenes where several subjects interact still need work[2], so keep one presenter and one simple action per shot.

Want the prompts and checks in one file?

The free prompt pack has the Seedance 2.5 template, three example prompts, the 25-second shot list and the quality checklist from this guide.

The workflow, step by step

Six steps, in the order we run them for every batch.

  1. Step 1

    Hooks and script first

    We write the hook before any generation: three to five openings per script, each a first line and a first image, then one script of 20 to 30 seconds. Every claim is one the brand can back up. The presenter shows the product and never speaks as a customer.

  2. Step 2

    Presenter and product references

    The presenter is a generated image or a licensed digital character: BytePlus will not take an upload of a real face unless that person has verified and authorized it[10], and CapCut restricted real faces when 2.0 arrived[6]. The product is a clean packshot with a sharp label. Keep both no larger than the output resolution, which BytePlus says reduces fingerprint-like textures[11].

  3. Step 3

    Image-to-video with Seedance

    Give Seedance both images, as references or one composed first frame, and generate one shot per row of the shot list. Seedance 2.5 can make 30 seconds in one pass[4], but a shot of 4 to 6 seconds is easier to check and cheaper to redo. We make several takes and keep one.

  4. Step 4

    Voice and lip sync

    Seedance makes the sound with the picture. Put dialogue in curly brackets and name the language first; English is one of the 11 languages listed[4]. If the lips drift, regenerate the shot, or swap in a text-to-speech voice and re-sync the mouth with a lip-sync tool.

  5. Step 5

    Edit, captions and placements

    Assemble the shots in a video editor with music, captions timed to each word and the on-screen text. Cut a 9:16 master, then 1:1 and 4:5, with text and product inside each safe zone. Our own video pipeline does captions, music and safe-zone checks in one command, and has made 30+ short AI explainer videos.

  6. Step 6

    Quality review

    Watch every shot at full size, frame by frame where hands and labels appear, then the whole cut on a phone with the sound on and off. What fails the checklist is regenerated or dropped, never shipped to make the count. Then switch on the AI label.

A prompt structure for UGC shots

BytePlus’s Seedance 2.5 tutorial sets the order: subject, action, scene, style, camera, then sound. It asks you to say what each reference image is for, and to put dialogue in { }, sound effects in < > and music in ( )[4]. Its prompt guide adds: describe what you want, keep negative instructions for subtitles and audio, and use timestamps in whole seconds, which 2.5 follows and 2.0 does not[11].

Template: one shot of a UGC-style ad, Seedance 2.5
Image 1 is the presenter: use the face, hair and clothes only.
Image 2 is the product: keep its shape, label and colours exactly.

[Subject, action] The presenter from Image 1 [one action]
with the product from Image 2.
[Scene] [One real place], [the light], [one or two props].
[Style] Handheld phone video, vertical 9:16, filmed at home.
[Camera] [Shot size], [static or one simple move].
[Timing] 0-3s: [the hook action]. 3-6s: [the demo action].
[Sound] English: {[the line, about 12 words]} <[one sound]>
No subtitles. No background music.

Three worked examples with generic products. The lines are illustrations, not client work.

Example: talking-head review format, a skincare serum
Image 1 is the presenter: use her face, hair and clothes only.
Image 2 is the serum bottle: keep its label and colours exactly.

The presenter from Image 1 sits at a bathroom counter and holds
the bottle from Image 2 to the camera, label facing out.
Soft morning window light. Handheld phone video, vertical 9:16.
Medium close-up, static.
0-2s: she lifts the bottle into frame and taps the label.
2-6s: she puts one drop on the back of her hand.
English: {Three things to know before you buy this serum.
First, it's fragrance-free.}
No subtitles. No background music.
Example: unboxing, a meal-kit box
Image 1 is the presenter: use his face, hair and clothes only.
Image 2 is the meal-kit box: keep its logo and print exactly.

The presenter from Image 1 opens the box from Image 2 on a
kitchen counter in warm evening light.
Handheld phone video, vertical 9:16.
0-3s: overhead shot, his hands pull the tab and lift the lid.
3-7s: medium shot, he holds a recipe card up to the camera.
English: {Here's what's in this week's box. Three dinners,
already portioned.} <cardboard tearing>
No subtitles. No background music.
Example: problem-then-fix demo, a phone stand
Image 1 is the presenter: use her face, hair and clothes only.
Image 2 is the phone stand: keep its shape and colour exactly.

The presenter from Image 1 sits at a home-office desk on a video
call, her phone leaning on a coffee mug. Daylight.
Handheld phone video, vertical 9:16.
0-2s: medium shot, the phone slides off the mug and she sighs.
2-6s: close-up, she sets it in the stand from Image 2 and
tilts it to eye level.
English: {Stop propping your phone on a mug. This folds flat.}
<soft click> No subtitles. No background music.

In all three the presenter states product facts. Nobody says “I have used this” or describes a result.

The same shot, written for Veo 3.1, Gemini Omni Flash and Kling 4.0, is in our AI UGC ads prompt templates.

A shot list for a 25-second ad

Six shots, each its own generation, so a bad hand costs one shot, not the ad. Seedance’s shortest generation is 4 seconds[4], so short shots are trimmed in the edit.

A shot list for a 25-second UGC-style ad
SecondsShotWhat it has to do
0 to 2The hook: the presenter mid-action, close, with the problem or product in frameStop the scroll. No logo, no greeting.
2 to 6The problem, shown in one simple actionOne idea, filmed as a person would.
6 to 12The product in use, in a medium shotLabel readable, grip believable.
12 to 17A close-up insert: the texture, the hinge or the packsMade from the product image alone, so the label stays sharp.
17 to 22The reason to buy, said onceA claim from the brief, with text added in the edit.
22 to 25The offer and call to action on an end cardBuilt in the edit, inside every safe zone.

Or have 4 finished ads made for $600

We run this workflow on every batch: the scripts, the takes, the edit, three placements with captions, the checks and the AI label. Delivered within 10 business days.

The quality checklist

On every shot before the edit, and again on the finished cut.

  • Hands. Five fingers, a believable grip, and no hand merging into the product.
  • Label and logo. The same words, colours and proportions as the packshot in every frame. Freeze the frame to read it: generated text can be misspelled[11].
  • Lip sync. The mouth lands on every word, above all on p, b and m, and the voice stays the same across shots.
  • Artefacts. BytePlus’s known issues: unwanted subtitles or music, ripple-like audio noise, fingerprint-like textures and glowing eyes[11]. Also faces, clothes or rooms that change between shots.
  • Claims on screen. Everything said or shown is in the brief and can be backed up. No before-and-after, no absolute claims[12], nothing that implies the presenter used the product.
  • Brand safety. No real face or voice without a signed release, no other brand’s logo or character, captions inside the safe zones, and the AI label on.

The rules most tutorials skip

As the platforms and regulators stated them on 15 November 2026. This is not legal advice: rules differ by country and category, so take anything regulated to your legal contact.

  • TikTok: label it. AI-generated content is allowed in ads with TikTok’s AIGC label or your own clear disclaimer, caption, watermark or sticker, and an undisclosed AI ad is rejected or restricted[12]. In Ads Manager, switch on “This ad contains AI-generated content”; Spark ads follow the community guidelines instead[13].
  • Meta: assume it can tell. Meta labels ads made or significantly edited with its own AI tools “AI info”, and says it will detect ads made with other AI tools through industry-standard signals[14]. Dreamina lists invisible watermarks and C2PA Content Credentials among its measures for Seedance 2.5[5].
  • New York: disclose a synthetic performer. Under a law signed on 11 December 2025, in force 180 days later, an advertiser who knows an ad contains a generated person who is not an identifiable real one must disclose it conspicuously. Fines are $1,000, then $5,000 for each later violation[15].
  • The FTC: a presenter is not a customer. 16 CFR Part 465 bars a business from creating or spreading a consumer testimonial that materially misrepresents that the person exists or used the product[16], and the FTC’s announcement names AI-generated fake reviews[17]. Its questions and answers say AI avatars are not banned outright, but a fake or false avatar testimonial is, and avatars could also breach the FTC Act[18]. Our presenters demonstrate and explain. They never claim to have used the product.
  • A real likeness needs a signed release. New York makes it a misdemeanour to use a living person’s name, portrait, picture, likeness or voice in an ad without their written consent[19]. TikTok bars likenesses used without permission[12], and BytePlus accepts a real face only once that person verifies and authorizes it[10]. We get the release before production.
  • Health and beauty: no generated “after”. TikTok bars before-and-after comparisons that may give a false impression of a product’s results, and its example is wrinkles disappearing after a cream[12]. Meta allows a before-and-after for general cosmetic products in ads targeted at people 18 or older, and bars statements of inferiority about appearance[20]. A generated “after” shows a result nobody had, so we never make one.

What it costs: a tool or a studio

On the API you pay by the second and the resolution. In Dreamina and CapCut you pay credits per second inside a monthly plan. Published rates as of 15 November 2026:

Seedance API price for one 5-second 16:9 clip without video input, BytePlus, as of 15 November 2026
One 5-second clip480p720p1080p
Seedance 2.5$0.514$1.156$2.843
Seedance 2.0$0.35$0.76$1.87

What moves the tool cost

BytePlus bills tokens: seconds times frame size times frame rate, so a 15-second take costs about three times a 5-second one. A reference video costs more, and only successful generations are billed[3], so a take with a bad hand is paid for.

Dreamina charges 37 credits a second for 2.5 at 720p, 17 for 2.0, 8 for 2.0 Fast and 5 for 2.0 Mini, with prices and allowances that differ by market[21].

A studio instead

Our trial is 4 finished ads for $600 and our pilot 12 for $1,800, each delivered within 10 business days, with retainers from $2,500 a month. That covers the scripts, the takes we discard, the edit, three placements with captions, the checks and the labels. The UGC ad cost page sets it against creators and tools.

Do it yourself for a handful of ads when someone has the hours and a good eye: Dreamina’s US Basic plan renews at $15 a month[21], a tenth of one ad in our pilot. A studio pays off when you need new ads every week and nobody has the hours.

Free download · PDF, 4 pages · no email needed

Seedance UGC ad prompt pack

The prompt template, three example prompts, the 25-second shot list, the quality checklist and the label and likeness rules from this guide.

Download it

Questions about Seedance UGC ads

Write the hooks and script first. Generate a presenter image, take a clean product photo, animate both with image-to-video one short shot at a time, and check the voice and lip sync. Then cut 9:16, 1:1 and 4:5 versions with captions, review every frame and apply the AI label.

Sources

Every outside fact on this page comes from the pages below, read on 15 November 2026. Our own prices are the ones published on our pricing page.

  1. ByteDance Seed: Seedance 2.0 official launch (12 February 2026). seed.bytedance.com/en/blog/official-launch-of-seedance-2-0. Read 15 November 2026.
  2. ByteDance Seed: One-take creation, flexible referencing: introducing Seedance 2.5 (31 July 2026). seed.bytedance.com/en/blog/one-take-creation-flexible-referencing-introducing-seedance-2-5. Read 15 November 2026.
  3. BytePlus ModelArk documentation: Model pricing, video generation models. docs.byteplus.com/en/docs/modelark/model-pricing. Read 15 November 2026.
  4. BytePlus ModelArk documentation: Dreamina Seedance 2.5 tutorial, with its comparison of the 2.5 and 2.0 series. docs.byteplus.com/en/docs/modelark/seedance-2-5. Read 15 November 2026.
  5. Dreamina: Giving creators more control with Dreamina Seedance 2.5 and Dola Seedream 5.0 Pro (23 October 2026). dreamina.capcut.com/resource/seedance-2-5-launch. Read 15 November 2026.
  6. CapCut newsroom: the Dreamina Seedance 2.0 announcement, with updates to 7 April 2026. www.capcut.com/newsroom/dreamina-seedance-2. Read 15 November 2026.
  7. CapCut: Seedance 2.5 tool page and plans. www.capcut.com/tools/seedance-2-5. Read 15 November 2026.
  8. Higgsfield: Seedance 2.5. higgsfield.ai/seedance/2.5. Read 15 November 2026.
  9. Replicate: bytedance/seedance-2.5. replicate.com/bytedance/seedance-2.5. Read 15 November 2026.
  10. BytePlus ModelArk documentation: Create portrait videos with Dreamina Seedance models. docs.byteplus.com/en/docs/modelark/seedance-portrait-asset-guide. Read 15 November 2026.
  11. BytePlus ModelArk documentation: Dreamina Seedance 2.5 prompt guide, with its FAQ of known issues. docs.byteplus.com/en/docs/modelark/seedance-2-5-prompt-guide. Read 15 November 2026.
  12. TikTok Advertising Policies: Misleading and false content (updated November 2026). ads.tiktok.com/help/article/tiktok-ads-policy-misleading-and-false-content. Read 15 November 2026.
  13. TikTok Ads Manager help: Add disclaimers to ads. ads.tiktok.com/help/article/add-disclaimers-to-ads. Read 15 November 2026.
  14. Meta: Expanding GenAI transparency for Meta's ads products (February 2025, updated 1 June 2026). about.fb.com/news/2025/02/gen-ai-transparency-metas-ads-products/. Read 15 November 2026.
  15. New York State Senate: Senate Bill S8420A, disclosure of synthetic performers in advertisements, signed 11 December 2025. www.nysenate.gov/legislation/bills/2025/S8420/amendment/A. Read 15 November 2026.
  16. eCFR: 16 CFR Part 465, Rule on the Use of Consumer Reviews and Testimonials. www.ecfr.gov/current/title-16/chapter-I/subchapter-D/part-465. Read 15 November 2026.
  17. Federal Trade Commission: Federal Trade Commission announces final rule banning fake reviews and testimonials (14 August 2024). www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials. Read 15 November 2026.
  18. Federal Trade Commission: Consumer Reviews and Testimonials Rule, questions and answers (November 2024). www.ftc.gov/business-guidance/resources/consumer-reviews-testimonials-rule-questions-answers. Read 15 November 2026.
  19. New York State Senate: Civil Rights Law section 50, right of privacy. www.nysenate.gov/legislation/laws/CVR/50. Read 15 November 2026.
  20. Meta Transparency Center: Advertising standards, health and wellness. transparency.meta.com/policies/ad-standards/restricted-goods-services/health-wellness. Read 15 November 2026.
  21. Dreamina: Seedance pricing (27 October 2026). dreamina.capcut.com/seedance/seedance-price. Read 15 November 2026.

Keep reading

Rather have the ads than the workflow?

A trial is 4 finished ads for $600 and the pilot 12 for $1,800, each delivered within 10 business days, with no commitment. Send a product link and the ads you run now, and we reply within one business day with the hooks we would test first.

Get an estimate